Bitget announced on April 6 that its AI trading agent, GetClaw, now operates from a dedicated account structure on the exchange. The account lets the agent hold assets, execute real trades based on natural language instructions, and manage positions in real time without a human account as the underlying holder, according to a Bitget press release via GlobeNewswire. The claims are self-reported and have not been independently verified.
The structural distinction is the announcement’s main point. Most exchange-connected AI trading systems operate via API access layered over a human account. A dedicated agent account means the agent holds the account directly, within parameters set by the user. Bitget frames this as moving AI from a proxy to a financial principal.
The Account Structure
According to the announcement, dedicated sub-accounts keep agent-driven assets separated from user-controlled assets. Users define strategies in natural language; GetClaw executes, monitors, and adjusts positions within those parameters without requiring manual intervention.
The agent account builds on two earlier Bitget launches: GetClaw itself, a zero-installation persistent trading agent, and Agent Hub, which added analytical AI Skills and real-time data tools feeding directly into execution.
“Sooner or later emerging financial markets are going to be filled with AI agents trading on behalf of users. We’re preparing the infrastructure to run this on scale,” said Gracy Chen, CEO at Bitget, according to the GlobeNewswire release.
Agents as Financial Principals
The account ownership question is one that any team building finance-adjacent agents will eventually hit. Whether an agent operates as a proxy on a human account versus as an account holder in its own right carries implications for liability, regulatory classification, and operational transparency.
Bitget’s implementation keeps ultimate control with the user: the agent operates within predefined parameters on a user-authorized sub-account. But the agent holds the account for execution purposes, not the human.
Bitget’s framing of this as an “agent-native exchange” is aspirational and self-described. Whether regulators, custodians, and institutional counterparties will treat agent accounts the same as human accounts remains an open question. This is a product decision by a single exchange, not an industry standard. But it represents a concrete infrastructure choice: financial infrastructure explicitly designed around agents as account holders, and a public statement from a major exchange that this is the direction markets are heading.